Audio level and Siglevdetcal optimization

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
Report Content as Inappropriate

Audio level and Siglevdetcal optimization

Mischa van Santen
Over the past two years I have developed some methods to ensure optimal audio dynamic and accurate adjustments for recieivers that are utilized within the Svxlink environment. These methods works fine for me but are rather time consuming and they partly require the use of a live system as a reference, which is undesired. Below I propose some ideas that will greatly reduce the complexness to properly adjust svxlink receiver(s).

---- Audio Level Calibration -----
When utilizing remote receivers, there is currently only one indicator in Svxlink that helps to determine the correct and optimal audio level, being the peak meter. Unfortunately this meter only indicates overload of the Svxlink audio processing, which usually is lower then your audiocard headroom. It would be very beneficial to have a calibration tool that helps to adjust the LF input and soundcard level in such way that an optimum dynamic can be achieved. Moreover it can help to ensure that multiple receivers are adjusted in a networked system with exacly the same "demodulation level" towards the svxlink core.

In our system, we currently utilize 17 receivers (and more to come). Calibrating all input levels equally is really a complicated job that requires sometimes disturbance in the live system because I simply need to adjust new receivers to the same levels with the live ones. A simple calibration addition in Svxlink could overcome this:

To be effective and accurate, an RF signal source is a necessity. Since not everybody owns one, it could of course be a transmitter with a defined tone (1kHz) and level, as long as it behaves stable over time to ensure that receivers are calibrated with the same levels. Drawback is that the exact deviation is then unknown; A real signal generator is therefore much more ideal. Dependant from the utilized channel bandwidth in an FM system, the signal generator should be set to e.g. 1kHz tone and a deviation that represents the peak value for that channel (2.5kHz in 12.5kHz bandwidth and 5kHz in a 25kHz bandwidth). The applied signal level should of course be such that absolutely no noise is present at the receiver.

A calibration tool should ideally show a bargraph or numeric indication of the receiver input audio level with a defined optimum, e.g. 100% or otherwise (I prefer dB scaling of some kind). This 100% should represent the maximum allowable audio level for svxlink at 1kHz. This ensures that the system is utilized to its maximum dynamic. By also indicating such numeric level with an adequate resolution, future receivers can be adjusted to that exact same level, ensuring that all receivers behave the same.

**note 1**
Adjusting the input level usually is a combination of setting the alsamixer levels together with adjusting the output level of the receiver in some or other way (e.g. potentiometer). Reason for this is that the alsa stepsize is dependant from the used audiocard, which is sometimes rather limited.

**note 2**
As svxlink or remotetrx works best with flat audio (due to the S/N measurement), it should be noted that the current maximum signal level is determined by opening the squelch ("hearing" noise) and then adjust the signal level just below distortion indication. With my method however, a defined signal (1kHz tone with appropiate deviation) is applied. The audio level of this signal will fundamentally be lower. The defined tone is required since different receivers will cause different audio frequency responses which makes it unsuitable to adjust audio levels at noise conditions. This is the prime reason to have such a adjustment process, rather then noise as some sort of reference. Ideally, such a level indication / tool includes a software bandfilter to eliminate residual noise and optimises comparison of receiver levels.

---- SigLevDetCal improvement ----
The siglevdetcal routine is a fine tool to calibrate the S/N indicator which is of high importance in diversity systems. Since all receivers differ in performance (sensitivity, SINAD), the S/N behavior will differ as well. You might be lucky if you utilize the same receiver (and same batch etc.), differences might be small. However, in larger systems it is barely possible to have exactly the same ones everywhere.

From my opinion it is therefore not adequate to only apply a strong signal into a transmitter while defining that as the 100% signal, compared to noise. The problem is that some receivers reach a 30dB S/N ratio at -70dBm while others reach this 30S/N ratio at -80dBm. In a diversity system, this easily leads to the effect that the voter switches (or remains) at a receiver that in fact has a significantly lower S/N than another one at that same time. Diversity performance will therefore often be poor, especially when many receivers are used.

Again, a signal source with adjustable level would be necessary to set a certain signal level that leads to a defined S/N ratio or SINAD value which then should be used for the 100% scale. This (25dB or 30dB) value should be the reference to all the (future) receivers in the system, ensuring proper voter switching.

For svxlink users who don't have the required measurement equipment available, the current SigLevDetCal routine suffices, but it is not ideal. Better would be to add a feature where in the first step a weighted measurement result (preferably 25 or 30dB SINAD) shall be achieved by adjusting the RF level of the source until the desired value is reached. This shall then be the basis for the 100% scale, followed by the current Siglevdetcal. In fact, a weighted measurement might even make the siglevdetcal routine redundant...

In our system, I run such a reference measurement in advance of the siglevdetcal with the aid of a CMTA84 radio testset (which holds a SINAD measurement function). The absolute accuracy of the SINAD weighted measurement is not that important within certain limits; it is only required to ensure that the reference for the 100% scale is the same at all times for calibration for each individual receiver.

Hope this helps you gents out ther and it probably greatly improves the audio quality and diversity behavior of svxlink systems (unfortunately I've heard many that can use a lot of improvement). The proposed additions in Svxlink would be a strong improvement from my point of view. I'd be happy to discuss the best method at all times.

Vy 73, 
Mischa - PA1OKZ

Transform Data into Opportunity.
Accelerate data analysis in your applications with
Intel Data Analytics Acceleration Library.
Click to learn more.
Svxlink-devel mailing list
[hidden email]